95. Emily Finn: Neural fingerprinting, 'naturalistic' stimuli, and taking time before starting a PhD

Emily Finn is an assistant professor at Dartmouth College. We talk about her research on neural fingerprinting, naturalistic stimuli, how Emily got into science, the year she spent in Peru before her PhD, advice for writing well, and much more.

There are occasional (minor) audio disturbances when Emily's speaking. Sorry about that, still trying to figure out where they came from so that it won't happen again.

BJKS Podcast is a podcast about neuroscience, psychology, and anything vaguely related, hosted by Benjamin James Kuper-Smith.

Support the show: https://geni.us/bjks-patreon

Timestamps
0:00:00: Supportive peer review
0:03:25: Why study linguistics?
0:11:05: Uncertainties about doing a PhD/taking time off
0:18:05: Emily's year-and-a-half in Peru
0:25:17: Emily's PhD
0:29:34: Neural fingerprints
0:49:25: Naturalistic stimuli in neuroimaging
1:24:01: How to write good scientific articles
1:30:55: A book or paper more people should read
1:34:58: Something Emily wishes she'd learnt sooner
1:39:20: Advice for PhD students/postdocs

Podcast links

• Website: https://geni.us/bjks-pod
• Twitter: https://geni.us/bjks-pod-twt

Emily's links

• Website: https://geni.us/finn-web
• Google Scholar: https://geni.us/finn-scholar
• Twitter: https://geni.us/finn-twt

Ben's links

• Website: https://geni.us/bjks-web
• Google Scholar: https://geni.us/bjks-scholar
• Twitter: https://geni.us/bjks-twt

References and links

Episode w/ Nachum Ulanovsky: https://geni.us/bjks-ulanovsky

Byrge & Kennedy (2019). High-accuracy individual identification using a “thin slice” of the functional connectome. Network Neuroscience.
Burkeman (2021). Four thousand weeks: Time management for mortals.
Finn, ... & Constable (2014). Disruption of functional networks in dyslexia: a whole-brain, data-driven analysis of connectivity. Biological psychiatry.
Finn, Shen, ... & Constable (2015). Functional connectome fingerprinting: identifying individuals using patterns of brain connectivity. Nature Neuroscience.
Finn, ... & Constable (2018). Trait paranoia shapes inter-subject synchrony in brain activity during an ambiguous social narrative. Nature Communications.
Finn, ... & Bandettini (2020). Idiosynchrony: From shared responses to individual differences during naturalistic neuroimaging. NeuroImage.
Finn & Bandettini (2021). Movie-watching outperforms rest for functional connectivity-based prediction of behavior. NeuroImage.
Finn (2021). Is it time to put rest to rest?. Trends in cognitive sciences.
Finn & Rosenberg (2021). Beyond fingerprinting: Choosing predictive connectomes over reliable connectomes. NeuroImage.
Grall & Finn (2022). Leveraging the power of media to drive cognition: A media-informed approach to naturalistic neuroscience. Social Cognitive and Affective Neuroscience.
Hasson, ... & Malach (2004). Intersubject synchronization of cortical activity during natural vision. Science.
Hedge, Powell & Sumner (2018). The reliability paradox: Why robust cognitive tasks do not produce reliable individual differences. Behavior research methods.
Sava-Segal, ... & Finn (2023). Individual differences in neural event segmentation of continuous experiences. Cerebral Cortex.

Chapters

• 0:00 Supportive peer review
• 3:25 Why study linguistics?
• 11:05 Uncertainties about doing a PhD/taking time off
• 18:05 Emily's year-and-a-half in Peru
• 25:17 Emily's PhD
• 29:34 Neural fingerprints
• 49:25 Naturalistic stimuli in neuroimaging
• 1:24:01 How to write good scientific articles
• 1:30:55 A book or paper more people should read
• 1:34:58 Something Emily wishes she'd learnt sooner
• 1:39:20 Advice for PhD students/postdocs

Transcript

[This is an automated transcript with many errors]

Benjamin James Kuper-Smith: [00:00:00] Yeah, I mean, I guess we'll be talking about, mainly about your work on neural fingerprinting and naturalistic, uh, stimuli or task paradigms. Uh, kind of before we do that, I, I thought I'd ask a little bit about something you mentioned, uh, you tweeted about a few days ago, which, which is, uh, which I thought was kind of interesting and I'd be curious for you to elaborate a little bit on.

So you said, just got reviews back on a grant. It's still a funding limbo, but one reviewer wrote such beautifully supportive comments of me and my work that I'm tearing up at my desk. I don't know who this reviewer is, but if you're reading this, thank you. This encouragement means so much. Uh, so yeah, I'm just curious if you could expand a little bit on that.

That's, uh, not an opinion you see people tweet about that much. Usually

Emily Finn: Yeah, it was, it was a rare personal, more personal, it's still work related, but sort of a rare personal tweet for me. Yeah, I submitted a grant last summer, as one does, and waited, you know, six long months [00:01:00] to hear anything, hear any news, and you know, you finally get this email, and it's like, well We're feeling positive, but we still don't know if we can fund it.

And so to start reading, there were five reviewers, five reviews and, um, start reading through them. And, uh, you know, everyone has constructive criticism as they do, um, some nice comments to you, but, but there was sort of one paragraph in particular that stood out where. One of the reviewers just wrote this long paragraph to saying, you know, that he thought that me personally, and the type of work that I was doing was really novel and innovative.

And, you know, it was really pushing the field forward. And, uh, he sort of said, you know, I'm going to go on to provide constructive comments because that's what we do. But to be very clear, you know, I have a very high level of enthusiasm for this investigator and this proposal, and I don't think she needs to, you know, revise it and wait another year.

I think you should just give her the money because. This is a really cool direction and she has this proven track record. And yeah, as someone who's only been in this PI role [00:02:00] for a few years and, um, you never quite know how people perceive you and how your work is landing. And obviously I've been fortunate to have some papers that I think have made an impact, but I don't know, just sort of getting that kind of personalized.

Encouragement from clearly someone who I'm assuming is a fairly senior person in the field, totally anonymously. Right? So they're not getting any credit necessarily for that from me or anyone else. Um, uh, other than this tweet, I suppose. Uh, and so, yeah, it was just sort of, it really moved me, to be honest, because, yeah, we, we work very hard on these grand proposals and most are not funded.

And sometimes you get lucky, but just to even get that piece of encouragement was, um. Was really lovely.

Benjamin James Kuper-Smith: Did that change a little bit how you're going to write grant proposal reviews yourself?

Emily Finn: I hope so. I hope that I will be in a position to pay it forward. I've been I've sat on a couple of grand panels. Um, not not too many so far, but I think it was a really nice reminder of, you know, when you really like someone's work, tell them, like, it really makes a difference. And, um, [00:03:00] I try to do that. And I have done that in the past.

But I think this recent experience was a good reminder and a good push of, you know, how much an impact that can make on someone. Um, because our field is filled with a lot of rejection and waiting and just, you know, kind of hard news. And so, yeah, I, I, I hope that I can continue to pay that forward.

Benjamin James Kuper-Smith: Yeah. Um I guess to get to your work and kind of what you do and What you've done so far, uh, I thought we could just take the kind of something that I do very frequently, which is a kind of biographical path of getting there. Uh, I'm curious, you had an article in the New York Times about you studying linguistics that I didn't read because I don't have access to the New York Times.

But I was just curious, I think I could see the first two lines where it says something you liked studying foreign languages or something like that. I'm curious, yeah, why did you decide to study linguistics?

Emily Finn: Yeah. Yeah. Um, I

Benjamin James Kuper-Smith: [00:04:00] Sorry, and just briefly, why did you, how did you get to write an article in the New York Times about how you study linguistics? That's slightly rather

Emily Finn: well, maybe I'll maybe I'll start with earlier than the article and kind of work my way up to the article. Um, so growing up, I'm pretty sure the first sentence of that article is growing up. I wasn't a science person or something like that. Um, maybe, maybe not. I don't know. You saw the first few sentences,

Benjamin James Kuper-Smith: something about how people in your, somewhere early on there was something about people in your school wanting to be like lawyers or something, or like

Emily Finn: yeah. Yeah.

Benjamin James Kuper-Smith: like that, yeah.

Emily Finn: Yeah. So, you know, I grew up, um, I went to a good public high school, um, and I was surrounded by, you know, lots of professionals and, but in high school, I was always very interested in languages and learning foreign languages. I kind of started with French and then my school opened up a new German program.

So I took a couple of years of German. I was kind of teaching myself Spanish on the side. I was just, you know, I really loved. Languages, um, which was not super typical in my high school, [00:05:00] but you know, there were good enough classes. So I was able to explore that. And then, uh, on my college applications, I didn't know exactly what I wanted to major in.

I kind of just wanted to take a smorgasbord of languages. I didn't even necessarily want to commit to one single language. So on my applications, I remember marking linguistics as my intended major, which of course here in the US, you know, It's infinitely, infinitely flexible, and you can switch majors. It wasn't like I was committing myself at the application stage to that, but I was indicating my academic interests, uh, and to me, the most natural thing was linguistics, because it seemed like it sort of gave me this opportunity to take some, to study some actual languages, but then also to learn more about how language works, you know, on a more fundamental level.

And so, um, I ended up going to Yale for my undergrad, and I did take some linguistics classes early on that kind of confirmed that, yes, I was interested in this subject matter. Uh, I did take one class called Language and Mind, which was sort of the cognitive science and a little bit of the neuroscience of language as well.

Um, and that really, really piqued my interest. Uh, and then I [00:06:00] ended up from there taking, uh, the sort of Intro to Neuroscience class that was offered at Yale at the time. And, uh, That was sort of have to fill a science requirement. So again, here in the US, you know, you have to kind of take a wide variety of courses that many of these liberal arts schools.

So I kind of needed a science class, but then I also kind of had an inkling from this other linguistics class that I might be interested in the brain. And, um, I really, really loved that class. And, uh, I kind of considered double majoring between linguistics and neurobiology, but I knew I never thought I wanted to go to medical school.

So, and the neurobiology major. Okay. Bye. Bye. Bye. at the time was, was kind of more of a pre medical. There were many classes they had to take that sort of overlapped with the pre med requirements, which I was less interested in. And so I ended up just sticking with the linguistics major, but I added a lot of coursework in neuroscience.

And I ended up doing a senior thesis in the linguistics department, but it was with a professor who was doing some fMRI of language. So this was an undergraduate senior thesis, um, but I was [00:07:00] very lucky in that this professor didn't have any graduate students at the time, so I was actually able to kind of start to take on the role of almost more like a graduate student with really helping to design and run this study.

It was an fMRI study of these kind of very esoteric, like, long syntactic clause sentences that, you know, are not something anyone would ever choose to utter, but nevertheless could be understood by people. And so we were sort of looking at, um, kind of pronoun movement and syntactic resolution of clauses in this sort of esoteric way, but at the time I loved

Benjamin James Kuper-Smith: I guess your, your, your interest in German must have helped a little bit there,

Emily Finn: Yeah,

Benjamin James Kuper-Smith: we're certainly

Emily Finn: in English, these sentences are like horrible, but I suppose in other

Benjamin James Kuper-Smith: it's alright. Yeah.

Emily Finn: Um, so I, I was doing this thesis my senior year, I was kind of sinking all of the free time I had into thinking about the study, writing the study, reading literature, uh, and then in parallel, I was, figuring out what I wanted [00:08:00] to do next after graduation.

So I was watching all my friends, you know, prepare their med school applications or go through the finance or consulting interviews. And I just really had no idea what I wanted to do. And I was kind of throwing a bunch of applications at different, of, of, um, of the world. Basically, you know, I was like looking at international fellowships.

I was looking at some nonprofit jobs. I was just kind of struggling to find anything that would feel like a fit and You know, in the meantime, I was, like, really loving my, my research, and I remember having a conversation with my undergrad advisor in the early spring of my senior year, when many of my friends had already kind of figured out what they were going to do, and I still didn't have any acceptances or any, really much to show for any of my efforts, and, and she was kind of like, well, you know, you really love this research, like, why don't you just go to grad school, and it had never occurred to me, uh, I'm sort of embarrassed to say, even after four years of college, that I could actually, you [00:09:00] know, you know, Become an academic and do this type of thing for a living, right?

And, um, yeah, I think, as I said, the growing up, I was surrounded by lots of, um, professional people that had wonderful careers, but I didn't know any professors or academics growing up. I think, um, even after again, making it through four years of college, I kind of had a pretty limited view of what professors actually did.

So I hadn't really thought about a PhD, but that comment really prompted me to introspect a lot and say, yeah, you know, this research is the thing that I'm. Really drawn to that. I actually enjoy the most. And so actually, that's when I wrote the New York Times. I say, I think the New York Times had put out a call that spring for, you know, essays from college students just about what was on their minds and how they were picturing and planning their lives after graduation and stuff like that.

So I drafted that and sent it in and It was very cool that they that they publish it. But yeah, the upside of the essay was sort of this research kind of snuck up on me or this career snuck up on me in the sense that it was kind of something I had been [00:10:00] doing, but I hadn't quite put it together in my mind that I could turn this into a career from there.

It was not necessarily a. A straight path from there. So I ended up sticking around for one more semester after graduation to wrap up the research or to, in theory, to wrap up the research I had been doing for my thesis, uh, as often happens in science, the honeymoon period kind of ended. So I think we got a bit lucky with the first couple of studies or analyses that we did for my thesis and things were working really well and I was, you know.

super excited with all the discoveries we were making. And then, and then of course, our follow up studies were a lot less clean and things got messy. And, um, yeah, I, I kind of, uh, saw the side of science where it's, it's hard because you're working really hard and you're not getting clear results and things aren't working the way you expect them to.

And, and so I had put in a couple of applications for graduate programs. I guess the fall after my after graduation, I was working as an RA and I was applying to grad [00:11:00] school, but kind of even by the time I submitted the applications. I was a little bit. I wasn't totally ready to jump into a program right away.

I think I was realizing that. Um, yes, I love this stuff, but maybe I should explore some other things, uh, before I

Benjamin James Kuper-Smith: What does not ready mean? Like you weren't sure about it, or you thought there was something you needed to learn before, or?

Emily Finn: I think I was feeling that, you know, this thing that I had really loved, like, maybe I only loved it because it had been going well. Like, maybe I just got lucky. And then once things started to get difficult, I was kind of like, oh, wow. Okay, this is, um. This is hard. Like, it's just emotionally hard. You know, you feel like you're working really hard and, and, um, things aren't working out the way you expect or hope.

And it's not clear if you'll have anything to show for your efforts. And so I, I was kind of going through that and seeing that side of things. And it made me realize like, okay, um, It's not just like you show up to work every day and [00:12:00] make a discovery and then you show up the next day and make another discovery.

You know, it's like these are long, winding, difficult roads and I mean that didn't totally scare me or put me off but I think it did make me realize that I needed to be really sure that this was the path I wanted to take in order to commit to At least, you know, five to six years of the PhD, uh, and then potentially beyond that, if I were to stay in academia, so I didn't think it was fair to either myself or my potential mentor, you know, for me to start a program, if I was experiencing this, this kind of uncertainty, uncertainty about whether I was genuinely ready to kind of throw myself into this.

Benjamin James Kuper-Smith: Would you recommend that to your, you know, you have undergraduate students now who probably ask you that kind of question? Because, I mean, in some sense it seems to me, It probably is a good idea to commit to the PhD period, but after that I guess it's, you know, you're not committing your life to, you're not signing off to a life in academia necessarily.[00:13:00] 

Uh, but I'm just curious, like, would you, um, because I guess many people are just unsure whether to do this or not, right? And in many ways you only find out by doing it, I guess. But I'm just curious, like, whether your perspective has changed on that a little bit or, yeah, just what you'd tell people.

Emily Finn: So I often give the advice that I've never met anyone who took time off and regretted it. I have met people who didn't take time off. And regret not taking time off, right? I've also met people that were really sure what they wanted to do and just went straight through and And they were fine and they thrived but i've also met people who went straight through because they weren't Quite sure what else they would do or because of the sense of you know These training periods are so long and I want to get what it doesn't have to be phd either I mean, it could be medical school.

It could be law school or, you know, whatever the next step you're considering, I think, um, especially when you come out of these kind of prestigious pressure cooker schools, there's this sense of like, okay, what are you doing? And, you know, you see all your [00:14:00] friends like going on to the next big thing, and it can be scary to kind of.

Take a pivot and do something for a couple years that you know is not what you want to do long term, but it just kind of gives you that breathing room to to figure it out. So, yeah, I mean, I, I often give advice, whether solicited or unsolicited to undergrads or other people that are kind of at a crossroads of, like, you know, If you do end up going on to a PhD, let's say, I mean, no one ever said, or I've not met anyone that said at the age of 40 or 50, like, oh, I wish I got my PhD two years earlier.

I mean, yes, it's a long road, but it doesn't really matter. And like, at the end of the day, um, like, I mean, it doesn't really matter if you take those extra couple years in terms of your overall lifetime line, but it does matter in the sense of if you take those couple years and it, yeah. Increases your confidence that this is actually what you want to do and that can kind of help sustain you through the difficult times of like, well, I tried a couple other things or, you know, I did take that time to kind of [00:15:00] explore and I came back to this, you know, it's kind of like that silly old saying, like, if you love something, let it go.

If it comes back, it's your, you know, it's like, it's like, if you do actually circle back to this, I think it's, that's very meaningful. And, um, and, and yeah, as I said, I've never really met anyone that. Took a few years off and regretted that. But I have met people who did not take, you know, a little time off when they had the opportunity and then later on said, you know, yeah, it may not have changed their ultimate path, but it might've just given them that confidence or those extra experiences to kind of power through when things do get tough.

Benjamin James Kuper-Smith: Yeah, it's interesting that you said that. No one's forget that because I guess I've sometimes voluntarily, sometimes not so voluntarily taken time off between each step. Uh, I mean, I went immediately to after school, immediately into my undergraduate, but then I had like, uh, kind of just happened that I did it one year working as a research assistant.

Um, I mean, I guess in Europe, you masters and PhD are separate anyway, usually. Um, so. [00:16:00] I had, I had a master's, but then summer job, they offered me to stay on longer and it's like, why not? And yeah, I definitely don't regret having stayed there and then done a bit of that. And then after my master's, yeah, so it's funny because it was very similar in the sense that I think I was, we were very good year with very good people, most of whom went on to do PhDs at very good places.

And I remember like when we were applying, I just felt this, I don't know whether the others also felt it, but there's this sense, like if you don't. figure out what you're going to do and where you're going to do PhD, like in the next few months, that's basically it. Obviously if we thought about it, we knew it wasn't, but it felt like kind of like you had to figure it out this, this time.

And then I ended up doing it, starting a PhD and then quit that after a few months because it just wasn't the right thing, even though it was supposedly on paper, very good PhD. And the funny thing is like, I mean, quitting a PhD is obviously also quite different from. Uh, just taking time in between because then it's a question [00:17:00] for future PhD supervisors.

Why did he quit? Is he going to do it again? That kind of stuff. But even now though, like I'm going to start a postdoc in a few months. I mean, first of all, nobody even knows I took a I mean, like, I'll tell them, I don't mind, but obviously I'm sharing this right now in the recording, but like, no one knows about it.

No one cares. A lot of people said they wish they'd quit actually. They just waited too long. But, uh, yeah, it's interesting because, yeah, as I said, I've, I've done, I've taken some pauses. Also a few months now after my PhD, I'm taking a few months pauses. I definitely, there's definitely nothing to regret about those, those few months here and there or a year.

Or in some, one case even two years, uh, of, of not doing that. So, I mean, I'm a little bit older now, but I don't care.

Emily Finn: Yeah, yeah, totally. Yeah, I think that's most people's experience. Um, yeah, so I guess to, to pick back up that thread I'm realizing I'm, you know, [00:18:00] proselytizing about taking some years off and I didn't even get to what I actually did. So I was applying. I was applying that cycle sort of right after I graduated and Even by the time I sort of press submit on the applications, I was waffling a little bit as to whether I was fully ready to jump in.

And so I ended up, um. Moving to Peru for about a year and a half. Uh, I, as you do, uh, I got a job working for a coffee company. This was really random, uh, but I had, I hadn't studied abroad in college, which was a bit ironic given my love of an interest in languages. And so living abroad was, um, Something that I really wanted to do and experience.

And so what was supposed to be kind of just a six month, I think I was telling everybody, including myself that I'll just go for like six months, basically in between applying to grad school and starting grad school, but I got down there and now it was great. And I was having a blast and I did actually get into a couple of places that year, but I, I, it just became clear to me that I wasn't ready [00:19:00] to commit.

And so I tried to actually defer my acceptance from Yale because I was pretty sure that I wanted to go back to Yale for my PhD and they wouldn't let me defer. And I was already in Peru and I just said, you know, I wrote some very heartfelt emails, but I was just like, you know, I'm not ready. And, um, so I'm gonna decline and, and thank you so much.

I'm so honored for this opportunity, but I don't think it's the right move for me or anyone else, you know, for me to start a program right now. And so I stayed in Peru for about a year and a half. Um, Had some really fantastic experiences. Uh, and then I ended up kind of starting to miss science. Right. So I, um, I ha I did had, I had had interests in kind of international development and, and, uh, and actually journalism as well.

And I was able to explore a little bit of the development side in Peru. But I did start to miss science, and I never knew, or I never thought I was going to be there forever, so I started to think, well, maybe I want to explore science journalism, maybe I want to [00:20:00] do science communication, and so when I first moved back to the U.

S., I was working at MIT in their news office, and I was writing kind of press releases or kind of public facing articles for their homepage, and that was really fun, so I got to interview, you know, a bunch of Amazing scientists at MIT when they would have big papers come out, you know, our office would do a big press release an article on it.

And so I got to kind of flex those muscles of, you know, taking interesting scientific findings and trying to communicate them to a lay audience. And I always like to write. And so that was, um, really fun. Like, I had a lot. uh, had a lot of fun in that role, I think. Um, but then part of it ended up being kind of this gateway drug back into science itself, because at some point I realized like I don't want to be just writing about science, I want to be the one doing the science, um, and especially when I would interview people in their BCS, their Brain and Cognitive Sciences program.

About their work, I just would want to, like, stay in those conversations, just keep talking about what was happening. And so, and that kind of [00:21:00] showed me, like, okay, maybe I am ready. Maybe. And, and, you know, nothing I had been doing, even though the work improve is really interesting. And then the work at MIT was also quite interesting, like, nothing that I was doing sort of on a day to day basis.

And either of those roles really captured my. Um, attention or passion or interest the same way that that research had. Um, and so that was a good sign that, okay, you know, I've sort of explored these counterfactuals, so to speak. I've kind of tried these other careers that I thought might appeal to me, and, um, I liked them, but I didn't love them.

You know, I didn't love them as much as the research that I've been doing, and so that gave me the confidence that, um, maybe I was ready to go back, uh, and commit to the PhD. I don't think at that time I was necessarily thought I was committing to academia for the rest of my life, but it was, you know, even just a five or six year, you know, here in the US, we don't often do masters.

Some people do, but, but, you know, it's possible to go straight into a PhD and then it takes five or six years. And that's a chunk of time. So, you know, you want to be sure that you're ready, even just for that step. Um, but I did feel ready at that point. Uh, and [00:22:00] so I, Reapplied, um, just to Yale. I was lucky that they took me back.

Uh, Todd Constable took me back. Um, and kind of the rest is history, I guess.

Benjamin James Kuper-Smith: Yeah, just a kind of last question about, uh, uh, the, the, the time kind of before you started, I guess, doing what you're doing now. Yeah. How did the thing with Peru happen? I mean, like, why Peru? How does one. I mean, did you, yeah, have a specific language or country or region or, uh, and then, I mean, you said you worked for a coffee company, which, uh, is pretty generic, I'm not exactly sure what that's supposed to mean.

Um, so what did you actually do when you were there and kind of why that job or, yeah.

Emily Finn: I wanted to go somewhere Spanish speaking. I never formally studied Spanish, but it's a very common and wide spoken language, both here in the U. S. and all over the world. So it was a language that I wanted to to learn and kind of cement for myself. So I wanted to be ideally somewhere in Central or South [00:23:00] America and the opportunity was, you know, I found it like on a.

Pretty random. There used to be a website called idealist. org, which was a kind of nonprofit jobs and stuff. So the company that I worked for was, was an interesting one in that it was technically a for profit coffee company headquartered in the U S in Portland, Oregon, but it, it was one of these kind of social enterprise models where.

They worked very closely with coffee farmers in coffee growing countries to improve the quality of their product, which can vastly increase the price that they can sell it for in these markets in the US and Europe. So the headquarters again was in Portland, Oregon, but they had these. supply side offices, one in Mexico, one in Lima and one in Tanzania.

And, uh, so the model was to kind of apply for grants or, or pour some of their own profits back into helping coffee farmers improve their equipment and their practices to kind of get the highest quality coffee, um, which you can then sell for, you know, [00:24:00] many, many times the price of, you know, a typical.

Batch of coffee just on like the New York stock exchange or whatever. So, uh, it was, it was a very interesting business model. I mean, actually what I was actually doing for them was, was a lot of communications and grant writing. And, um, but I was based in their Lima office and I got to travel a bit to the coffee farms in Peru.

Uh, and it was really fun and really interesting. Uh, I can't say I use a lot of that knowledge at this point, other than, you know, I drink a lot of coffee to fuel my neuroscience habit, but, um, so that, yeah. That was it.

Benjamin James Kuper-Smith: Okay, I don't know, it sounds fun, especially if you're into coffee.

Emily Finn: I, I, the ironic thing is I did not actually drink coffee before I worked for that company, but now I'm, I'm pretty addicted and I wish I could say I have like great taste in coffee because the, this company was very much about like high quality coffee and getting the best possible, um, beans from, you know, from the trees into your cup and every step along that process, like can have an effect on the final taste of the coffee.

In reality, I just [00:25:00] drink whatever. I mean, I can appreciate good coffee, but I'm not a snob. I just need any coffee. That's fine. Okay.

Benjamin James Kuper-Smith: Maybe that's, I also don't drink coffee. I just never like the taste of it. Uh, maybe that's how I can get into it too. Go to Peru and work there for a few years. Uh, anyway, we've kind of reached the point where you start doing research in a more full time, more committed sense. Yeah, you said that Todd Constable let you in again or something like that, I think was your phrase.

Um, so was it clear from the beginning that you were going to work with him? And if so, why?

Emily Finn: I think it was clear to me in the sense that I think I'd had enough experience by that point to realize that who your mentor is and who your advisor is has a big impact on, you know, not only what what's happening in your PhD but also beyond and so I'd gotten to know Todd, a bit in undergrad because my even though my primary advisor was in linguistics department.

We were running an fMRI study. [00:26:00] And it was her first fMRI study at Yale. So we were relying very heavily on Todd, who was at the time and still is kind of the director of MRI research at Yale. And so he, I got to know him quite well through that experience and other members of his lab. And yeah, I think at that point I was, I was ready to commit to the PhD, but I really wanted to know that it was going to be an environment.

That I could work well in and be happy and with people that I could work well with. And that was kind of a known quantity for me. So I didn't explore too heavily, um, beyond that, uh, that environment. And, and yeah, so,

Benjamin James Kuper-Smith: And he didn't mind you deferring and then coming back again?

Emily Finn: yeah, I mean, I think that was another benefit in retrospect, cause now, I mean, I, I'm sort of. I encourage people to take non traditional paths. I'm very open to students that have gone and done stuff outside, but, but now sort of looking at the, the path to PhD, at least here in the U. S., it's like, you know, usually you do a bunch of research in undergrad, and then you spend two years working as a lab manager or a research assistant in another lab, and then you [00:27:00] apply to grad school, and, and so you have sort of these experiences under your belt.

And, you know, I wasn't coming in with that full time RA experience right after college, and, and so I guess that made me a bit. non traditional, but I guess Todd was willing to take a chance on me, so to speak, because he'd seen me, uh, or he'd worked with me when I was doing my undergraduate thesis, and that was kind of evidence enough for him that, you know, I, if I was coming back, that I was serious and, you know, I was the real deal and I could get stuff done.

So yeah, I'm not even sure anyone else would have even looked at my application given that I'd sort of taken this weird detour. Maybe they would have, but, um, it just seemed like, like a good fit.

Benjamin James Kuper-Smith: I mean, that's the question, right? Because in many ways it's also just way more interesting someone who does something unusual and isn't too Afraid to, you know, I mean, in a sense, like I'd rather have someone, I mean, I'm not a PI, so who knows what would happen if I, if I were a PI, if my opinion would change, but I feel like I'd rather like to have people who are a bit interesting and not just [00:28:00] went straight down the path that was laid in front of them.

I mean, it's fine if you do that, but.

Emily Finn: I agree with you, but I also recognize that it is a bit of a risk, right? Taking one of those students. People have varying levels of openness to that, but Todd, I think, was secure enough at the time and, and, and kind of had some, some data on me, so to speak, for before and, and was willing to give me this opportunity.

So.

Benjamin James Kuper-Smith: Okay. And then, um, I don't actually know what your. Should probably look this up. Uh, what your initial projects were. What was it? It wasn't language again, right?

Emily Finn: Actually, funnily enough, my very first project in grad school was on dyslexia, which is not something I had studied necessarily in my linguistics major in undergrad, but obviously language related. And so they had an existing data set in collaboration with some dyslexia researchers, the Shaywitz's at Yale, very well known dyslexia researchers, and they had scanned like, I don't know, something like 140 kids and adolescents with dyslexia.

So [00:29:00] my very first project was looking at Differences in functional connectivity between, uh, typical readers and, uh, readers with dyslexia. Uh, and, and that was good. I mean, it was, it was an interesting project. We had some interesting findings. I think it, it served to, uh, kind of help me get my feet wet with the functional connectivity methods, uh, at the time, which were, which was what Todd's lab was interested in and ended up forming the bulk of my PhD.

So that was my very first project. And then from there, we kind of moved into the more. Individual differences and behavioral prediction, fingerprinting stuff.

Benjamin James Kuper-Smith: Yeah, I wanted to, I guess we can just move straight to the, to the finger printing stuff. Uh, now, yeah, to kind of, in a sense, introduce the, the topic. Um, I mean, I guess, I guess no one likes justifying their research, but I am going to ask a little bit like what, what the, why to do that. And I, um, so first you, you, I think you summarized this very well later in a review in your image.

Uh, you wrote, [00:30:00] but why do we care about individual differences in neuroimaging measures in the first place? Most researchers are probably not interested in brain based fingerprinting for its own sake. After all, there are better ways to identify someone than going to the trouble to scan them and calculate a brain connectivity profile, e.

g DNA, actual fingerprints, simply looking at or speaking to them. Uh, and to that I'd maybe add So I don't actually know whether this is true, but it seems to me that kind of the uniqueness of a human fingerprint is It's kind of just a random occurrence. I don't know whether there's any like, meaningful, but I don't think there's any meaning, I think it's kind of just, it just kind of happens to be that way and it happens that that is very useful in some criminological circumstances.

Um, so kind of what, what is the kind of specific application of neurofingerprinting that's kind of, I hope you can comment a little bit on that too. Yeah, so we know why we're, why we're even talking about this in the first place.

Emily Finn: This is a great question, and maybe to answer this, I'll kind of back up a step and say that the [00:31:00] fingerprinting result in that paper. So the 2015 paper really has two sides to it. The first side is we can fingerprint or quote unquote identify someone based on Their brain connectivity across different days or across different scan types, etc.

So that's sort of a brain to brain matching. You know, we're matching one person's brain on in one session to the same person's brain in a different session. The other half of that paper is the behavioral prediction side. So, can we actually take that brain data and use it to predict something about behavior outside of the scanner.

So that's sort of brain to behavior as opposed to just brain to brain. The fingerprinting side of that paper was very much a happy accident, so it was not the result. I mean, this, you know, when we tell our post hoc story about it, this is not the story we tell in talks, but this is the real story. So we at the time, so it was myself and Sheila and Shen, who is a fantastic research scientist in Todd's lab at the time, who were actually co first authors on that paper.

And, um, we had gotten [00:32:00] excited about kind of dynamic functional connectivity and also So. Differences in functional connectivity between brain states. So for example, between rest or performing different tasks, which, um, was popular at the time. And so we were playing around with data from human connectome project, and we had calculated these sort of sliding window connectivity matrices, and we can get into what that means.

But essentially we were sort of taking snapshots from, um, different individuals. And in that data set, the human connectome project, there's, um, many individuals who were all. Uh, scanned while doing different things. So each individual had two resting state scans and then, um, several different tasks scans, like, for example, a working memory and back task and emotional faces task, relational matching task.

And so we had these. Uh, snapshots, so to speak, of each individual under these different conditions. Uh, and what Shilan and I, uh, first set out to do was say, Okay, if we have a bunch of, uh, these functional connectivity matrices calculated from different tasks, [00:33:00] uh, from different people, and we throw those into a clustering algorithm, for example, can we cluster connectivity based on task state?

Can we see differences, reliable differences, between What the brain looks like as it's doing an end back task versus, you know, doing an emotional faces task based on on on the whole brain functional connectivity patterns. And so Again, we had this sort of set of matrices coming from different individuals under different conditions.

And every time we would throw them into these clustering algorithms, we kept getting out subject instead of state. So all of the matrices from the same subject would cluster together. So like your matrix from rest would look more like your matrix from And back than my matrix from rest, for example. And so, like, as it often happens at first, you know, we're sort of showing these results and everyone's like, well, that can't be right.

I mean, because the usual model in fMRI is, you know, you give a task and you think that that's that's kind of your manipulation of brain activity, right? Like brain activity changes and responses to tasks like we [00:34:00] know this. This is why fMRI works. Um, so why aren't we seeing, you know, consistent task patterns?

Uh, why do we keep getting. Um, Subject out, but then that result was so strong and we, we couldn't kill it. So we ended up just kind of leaning into that result. And that ended up being the finding, which was that actually there's a lot of stability within subjects and also uniqueness across subjects to these whole brain functional connectivity patterns, such that you always kind of look most like yourself, regardless of what you're doing.

And yes, tasks do change and modulate the connectivity on top of that, but they don't do so, so much as to mix. Me look like you, for example, or vice versa. There's some sort of stable backbone that's always present in my brain that yes can be modulated slightly depending on what I'm doing, but it's strong enough to kind of identify me across these different contexts and days.

It wasn't the result that we set out. It's not what we hypothesized, but it's what we found. And so I think it was a nice example of kind of [00:35:00] following the data and letting the data tell us, you know, what the most important thing was. And then we started to kind of formalize that angle of things and kind of poke that in different ways and see if we could make an interesting point about that.

And. It's funny because, yeah, I mean, as the quote that you just read kind of highlighted, I mean, to me, it always felt like, like a bit of a party trick of like, oh, look, we can identify people using functional connectivity. And, and we thought it was a compelling point, or we thought it would be compelling for fMRI researchers in particular, right?

Because there's a lot of kind of the traditional dogma of fMRI is like, yes, it works, but it's pretty noisy. And therefore, you need to aggregate data over multiple individuals to kind of find the effect that you're looking for, right? And this was kind of turning that on its head a bit and saying, actually, there's a lot of individual differences that are, yeah, they can be somewhat noisy, but they're stable enough within a person that you can actually look at, you know, this individual person's data.

Yeah. Uh, and that is the signal in that is sort of strong and consistent enough across. [00:36:00] different sessions or different brain states to uniquely identify that person. And so we thought that that would be compelling for the fMRI community. We didn't necessarily expect it to be as compelling as it was to like a more general audience.

because of exactly the, this point that you just read out, which is kind of, well, we don't need a method of identifying people based on fMRI. It's more like this is a methodological point we can make about, you know, there's enough stability and enough uniqueness in these signals, um, that they could potentially be meaningful.

And so for me, the, the more compelling piece back then, and, and, and still now is what do these fingerprints, so to speak, What kind of information do they actually carry about people? Because, uh, the analogy that I, I like to give, which I think picks up on what you were just saying is, you know, it could be the case that these functional connectivity profiles are just like barcodes, right?

Like if you go to a store and [00:37:00] there's, and you are looking at objects, like each one has a unique barcode that identifies it, but there's nothing in the pattern of thick or thin stripes, you know, for that particular object that tells you anything meaningful about the object. Like if you were just. to see that barcode in isolation, um, it wouldn't index anything meaningful about what that object is.

That could have been the case with these fingerprints. We were hoping that it wasn't, right? We were hoping that there was some meaningful signal in there that related to um, behavior or the actual real world output of the brain, right? And so, um, the second half of that paper was, was trying to prove to ourselves and others that, you know, there is meaning in these things.

And so that's where, um, we ended up, uh, showing. At least in that first paper that you can use resting state connectivity patterns to predict, to some degree, not, not a perfect degree, but with some degree of statistical accuracy, uh, predict people's levels of fluid intelligence, um, based on a test that they had done outside of the scanner.

And so that was kind of the piece that. I think we needed again to prove to ourselves and others that these were not [00:38:00] just sort of in fact that that kind of makes the fingerprint metaphor breakdown, right? I mean, the fingerprint metaphor is useful to kind of quickly convey what it is we're doing. But, but I fully agree with you that, you know, the actual pattern of bumps and ridges and grooves on people's fingers generally is, you know, not.

Maybe there's some, some pseudoscience that analyzes fingerprint patterns, but I think generally it's accepted that that doesn't actually index anything meaningful about a person. Um, uh, and that's in contrast to brain functional connectivity fingerprints, which we do have a fair amount of evidence now, like, can be used to index meaningful things about people.

Benjamin James Kuper-Smith: I guess there probably would be, right? I mean, if people do like palm readings, I guess the fingerprint is just too small to really see. So I guess

Emily Finn: I'm sure there's, I, yeah, I'm

Benjamin James Kuper-Smith: but if it was a bit bigger than that, right.

Emily Finn: Maybe. Yeah.

Benjamin James Kuper-Smith: Um, I don't know how relevant this question exactly is, but it just occurred to me.

I thought I'd ask it. Um, does this have any, uh, ethical implications for data sharing? I mean, in a [00:39:00] way, you know, sharing fMRI data, in a way it feels to me, it wouldn't because I guess you're already sharing like lots of information about people, but I'm just curious, like whether that has thrown, yeah, just thrown up any kind of particular implications there for you or how you share your data and that kind of stuff.

Emily Finn: It's funny because when the fingerprinting paper first came out, we did get a fair amount of media attention, which, you know, I was a grad student at the time and it, it, it seemed really fun in the beginning. And then it was actually quite stressful because we would do these interviews with journalists who would actually, you know, be sort of pressing us to admit that this has some privacy concern.

And, and I don't want to entirely dismiss that. I mean, I think, um, you know, any amount of medical data, uh, you have to think about, about privacy. I think, I mean, it's, you're still much. That are able to identify someone based on their anatomical like T one scan, for example, uh, and so when we share those, we often deface them.

And there's kind of ways around that. And of course, you know, anytime you're sharing [00:40:00] data, you want it to be very much anonymized and identified. And, um, there's no way to link it back to the individual. I mean, in theory, right? With with functional connectivity fingerprinting, if you had a scan from someone and you knew their identity, and then even if there's other scans of theirs kind of floating out there that have been identified, you could.

Potentially link it back to to this scan. I mean, I think the extent to which that poses a privacy concern really kind of depends on where we're going with this. Like, is it actually possible to get? I mean, I sort of said this before. I mean, we can get some degree of prediction accuracy for this fluid intelligence metric, but it's nowhere near ready to be rolled out in like a clinical or real world setting.

You know what I mean? And so I think for me, to some extent, it's like, well, Whether how much of a concern this poses really depends on how good our models will get because it's it's hard to imagine a world where I mean, unless you could meaningfully extract additional information [00:41:00] from someone's scan, just kind of knowing like, okay, this scan came from this person's brain.

And again, I don't want to dismiss it because I think many people are working towards that. And that's sort of like our ultimate goal or hope is that, you know, these things will have real world relevance. But I think. Okay. Right now. I mean, there's also the compliance piece of like, you have to agree to an MRI scan and lie actually quite still and be very compliant, um, for for a decent chunk of time in order to even get high quality enough data to do this.

Right? So it's not like we can just pick up on fingerprints that people have left behind on objects unwittingly. I mean, this is something that people, you know, volunteer for these studies and then we're quite careful with how we. Um, anonymize and identify the data. So I, uh, yeah, again, I don't want to tirely entirely poopoo.

And I think it's a thing that we have to be thinking about. But I think ultimately, if the goal is to use these scans to develop, you know, real world tools that are useful, um, then hopefully we can kind of do that in a way that respects people's [00:42:00] privacy, but, but also, you know, leads to some kind of, uh, meaningful clinical tool that that could be helpful.

So, um, Yeah,

Benjamin James Kuper-Smith: Yeah, I mean, I guess the, the, the, the point is that the main thing you're able to. I mean, for example, you know, as you said, like you can identify people across tasks, but it's already in the data that this is the same person, right? So it's not really from, from that immediate case, you're not getting that much.

Now, one thing that just occurred to me, can you, you know, how like often you have the same participants coming over and over again at like certain scanners to take part? I'm just curious, do you think you could, let's say you take like the fMRI papers that came out of one institute? And just figure out like, actually you only have like, you know, you say you test it, you know, you tested like a thousand people, but actually it was only like 300.

Emily Finn: again, it's like, it's hard to be fully sure because like most things like this is not perfect. And even when we're getting like 90 percent accuracy, there's still going to be some mistakes that we make. So I think it's [00:43:00] hard to like conclusively prove you could say, you know, oh, there's a chance, you know, my confidence interval is this only comes from, you know, half the subjects that saying it does.

But yeah, I think it'd be hard to like

Benjamin James Kuper-Smith: Yeah. And it's not like it would invalidate the data or anything like that. It was just curious, like whether you could, because I guess that would be like a, a way to use the general approach with existing data to see kind of how far it gets you all that kind of stuff. Um, but yeah, so maybe kind of the obvious question.

So what. What does identify people? Is it just some very nuanced statistical correlations between brain areas that are just, you know, if they're in the data, but as a human, you can't really, in that sense, do much with it? Or is it kind of fairly obvious that like, I think you mentioned, I think I saw talk of yours online, for example, that the social areas or something like that, or social stimuli?

Are particularly good for identifying people. So I'm just curious, um, I guess we're gonna move also more towards naturalistic stimuli and that kind of stuff now. Yeah, what inaught [00:44:00] decreases how easy it is to fingerprint someone? Neurally.

Emily Finn: Yeah, it's a good question. So in the first, in that first paper, we did a couple of of analyses where we kind of broke things down by brain network. And so we were asking, are there particular regions or networks of the brain that contribute more to this identification that are basically sort of like more distinguishing of people?

And, uh, I want to be careful with how I present this result, because we did find something fairly logical, which is that, uh, the higher order association cortices, so things like frontoparietal area, uh, frontoparietal network or default mode network, things like that, which had been known, you know, from a lot of prior work to be, to have more variability, both in terms of their anatomy and their function across people, those were the same regions that, relatively speaking, did better at identifying people.

That being said, One thing I think is, is also really important to keep in mind is that, you know, identification was possible well above chance based on all of the networks that we [00:45:00] tested, you know, so, like, it might be a difference between, like, 85 and 90 percent accuracy, for example. So, yeah, you can say, like, you know, some, there's some regions or networks that carry slightly more of this information, but I do think the information is very widely distributed in the Connectome, uh, and there's been some.

really interesting follow up papers, not, not by us, but by folks like Dan Kennedy and, and other groups that have shown that even if you just sort of select a random set of edges or connections, maybe as few as like even just 40 connections I think they found in one paper that are just randomly selected from all over the brain, like that's enough to fingerprint people with a decent degree of accuracy.

So I, I do think that there is something, there are particular brain networks and also. Um, states, which we can get to, uh, next maybe, but that, that do like a bit better, for example, than, than like average, and, and that's important to pay attention to, but I think it's also important to emphasize that the information is very widely distributed, and you can get very good fingerprinting, even with only like a very small [00:46:00] slice of the functional connectome, so to speak, that's kind of randomly drawn from all over the brain.

I mean, I

Benjamin James Kuper-Smith: Does that tell you anything interesting?

Emily Finn: even things like.

Benjamin James Kuper-Smith: Like, I mean, in a way you would, I would imagine that, for example, something, you know. Like the brainstem or motor cortex, something like the, the things that have to connect to things outside of the brain, uh, very directly or like early or sensory cortices. I would imagine that those might be a bit more stereotypical in people.

Um, and that the kind of, the more abstract you get, the, the more it's down to just random chance of how things, you know, work out in a particular brain. Um, I'm just curious, like, is there anything there that, that tells you anything kind of fundamentally. interesting kind of about how brains are structured or is it just kind of what I said it's the kind of fairly obvious things.

Emily Finn: Motor cortex or subcortical brainstem areas, they don't do as well again as the higher order courtesies usually, but they, they still do do [00:47:00] quite well, right? In terms of fingerprinting. And so I think, um, I mean, part of that could be due to. Perhaps less functionally interesting reasons. Um, for example, 1 question that we got on that paper and we still get, which is a great question is, you know, how much of this is just anatomical differences between people that then when you do your.

registration of your data and then you're applying this sort of group level atlas, although you can also do this on individual level, like how much of that is just sort of anatomy, which we know is slightly different between people, right? And so then that's going to have an impact on your functional signals if you're kind of grouping them according to these anatomical regions that are not exactly the same in everybody.

Um, so I do think that Differences in anatomy contribute to this signal for sure, and we know that anatomy differs across individuals, even in these kind of lower order regions, so to speak, or kind of typically more stereotyped regions. But I don't think that's the whole story again, because we do see differences across, um, brain state and how [00:48:00] well we're able to identify people.

And anatomy is not changing across state, at least not at those short timescales. So I think there's, there's additive components of both anatomy and function that are at play. And I think the fact that, um, yeah, the fact that this information is so distributed, again, like probably some of it is due to anatomical differences that are present all over the brain that you can always pick up on.

Um, but I think, you know, even in some of these, systems that we do think of as more stereotyped and more consistent. I mean, there's still individual variability in those systems too, right? And, and so I think it's, it's, it's a combination of all of those things.

Benjamin James Kuper-Smith: Yeah yeah I mean you'd imagine that anatomical differences would play a fairly major role but yeah

Emily Finn: a couple of things to try to, you know, rule out that as a, as the only reason, um, We definitely, I think that that is part of the reason, but to me, the most compelling evidence against that being the only reason for these differences is the fact that we [00:49:00] do see sometimes pretty substantial differences in.

Either how well we're able to fingerprint people or how well we're able to predict behavior based on just what the person is doing during the scan, right? And anatomy is not changing within the course of a scan, you know, between arresting state run on a movie run, for example. So, um, the function is definitely adding something on top of that.

That's meaningful for the individual signals that we can get out.

Benjamin James Kuper-Smith: so so let's then move towards different kinds of paradigms you might use and how that Changing. Uh, you had a paper called, I believe, Is it time to put rest to rest? Um, in which you talk about three waves of, uh, I guess, types of stimuli that or tasks that people do when in a scanner. Uh, so maybe just as an introduction, kind of what are the, what are the three waves and then we can get into a bit more detail.

Emily Finn: Sure. So in that paper, I was conceptualizing of this is okay. So the first wave was, you know, if we go all the way back to the [00:50:00] early nineties, when, um, fMRI first came on the scene, people were using, I think, rightfully so, you know, pretty simple and standardized tasks where they had a very clear handle on the timing of the task.

So, you know, you're asking someone to tap their finger at a particular time and you know when that time is and therefore you can search in the motor cortex for the answer. activity that increases as a function of tapping your finger, right? And then, um, there's visual tasks, which are very well controlled, where you're presenting, uh, flashing checkerboards or, um, other stimuli to try to map out retinotopy, for example.

And you have very discrete trials, and you know exactly, you know, what the person is seeing and when, and the stimuli are simple enough that you can, model them appropriately. Uh, and that was sort of the whole general linear model or GLM approach, which was the bread and butter of the early days of fMRI, which again, makes a lot of sense.

It's a brand new tool. Um, you want to, uh, make sure that you're doing studies in a way that you have clear hypotheses and you can validate the methods and understand and trust the signals [00:51:00] that you're getting from this new bold contrast. Right. And then, so that was kind of the first wave. Was these well controlled, uh, very carefully designed tasks.

Then the field moved into what I called in that paper, the second wave, which was resting state, right? So, um, one of the first papers was brought up as well as paper in 2005. It took a few more years, uh, than that to really kind of catch on. But I would say like by the early 2010s, people were really moving, uh, really getting interested in these spontaneous fluctuations, which.

Could be measured while people did nothing at all and just sort of lay in the scanner. Well, they're doing something. We just don't know what, right? So people are doing something, but it's not necessarily being controlled by the experimenter, but we're able to measure, you know, we get these really high dimensional signals and we're able to look at spontaneous fluctuations and understand, use those to understand something about how the brain is functionally organized.

Um, I mean, one of the reasons people first even trusted resting scans in the first place is because A lot [00:52:00] of the co fluctuations recapitulated what we were seeing during tasks, right? So the left and right motor cortex, for example, tend to be active together, even while people are resting. Um, and that kind of mirrors, you know, what we see in in tasks or or the frontal parietal network, for example, which we had seen during classic working memory tasks.

both the frontal and the parietal parts of that network would sort of be active often, um, when people were doing working memory. And we can see the same thing, you know, those two discontiguous regions are also active together when people are just resting. So. That was kind of the second wave of, well, maybe we don't need this tightly controlled task structure and we can just get a lot of information from looking at the dynamics of those signals, um, as they arise more spontaneously.

And I think that was great. I mean, I think that was extremely productive. I mean, obviously, the proof is in the pudding. Um, there's been a lot of advances that we've made, I think, in our understanding of brain functional organization, just from observing the brain at rest. But the argument I was making in that paper was I personally think [00:53:00] it's time to take a lot of that knowledge and a lot of that creativity that we were spurred to by this resting state.

Cause resting state, it really puts a lot of the burden on the analysis, right? Like you have no experimental design. And so all of the creativity and all of kind of the intellectual burden is on, well, how are you going to analyze the data in a way that, that allows you to, to get something out of it and draw insights that are interesting and interpretable.

And so we, we gained a lot from that for sure. But my argument is that, so now we can sort of. Take that as inspiration and move into this third wave where we're not necessarily just having people Lying to Skinner with no idea of what they're doing And we're also not going all the way back to the more simple control tasks Although I do think you know There's still a place for both of those things But we're now at a place where we can really move forward into this third wave which is thinking about how are we gonna?

Come back to Experimental design where we do have more of a handle on what people are doing or might be doing and design new paradigms or come up with clever new ways [00:54:00] of taking the new analysis approaches that we. That we developed in the second wave of resting state and taking those into a place where now we can start to link some of those fluctuations to ongoing cognition or sort of ongoing thought processes, which we don't have a great handle on at rest.

And I would argue, we also don't have a great handle on when we're using very simple, uh, artificial feeling control tasks. So it's kind of a marrying of the task based approach, not going necessarily all the way back to the simple control tasks, but thinking about, um, kind of a new wave of tasks that. are closer to what the brain might be doing in everyday life, but that we as the experimenter still have some knowledge of and some control over.

So there's some ground truth as to what people might be feeling or experiencing in that moment. Uh, and then can we relate that to these very high dimensional signals that we're recording using fMRI, taking some of the analysis approaches that were inspired by the foray into REST.

Benjamin James Kuper-Smith: So just before I forget it, why is the term naturalistic stimuli not ideal?[00:55:00] 

Emily Finn: This is, this is a soapbox that I've tried to get on and I think it's not, I still use the term for the record, but, um, it's funny because, well, No, I mean, I think, uh, I, I like to bring up that point to remind people. So, uh, and just to define it, I guess, um, for folks. So, like, usually when we say natural stimuli, we mean things like watching movies or listening to stories in the scanner.

These are often stimuli that were not expressly designed for research purposes. So these are. You know, movies that were made for the purpose of entertainment or, um, education or something else, um, or podcasts that are, yeah, entertainment or, or instruction. Uh, and so these weren't crafted for scientific research purposes and nevertheless, they can be very fruitful to use in neuroscience and psychology experiments.

Uh, but I, you know, I think the idea that a movie is naturalistic, I mean, a movie is, very deliberately crafted and, you know, very, yeah, just, just deliberate and specific choices about [00:56:00] editing and, and, and, and script. And, um, uh, it's funny, like people often think of like, for example, um, luminance. So just the brightness of the movie at any given time, it's kind of this low level.

confound, right? Like you want to regress that out. But if you talk to a filmmaker, they're like, luminance is not, I mean, luminance is a very deliberate choice. You know, you make choices about the lighting, depending on the mood that you want to set. And so I think it's just important to remember that, you know, it's not like we Walk down the street and experience our lives as a Hollywood movie, right?

I mean, there's a lot of, um, editing and crafting that go into these stimuli that are designed to really, you know, have an effect on people. Um, and that can make them very powerful stimuli, but I don't think, you know, I think it's a mistake to think that, yeah, if we were to walk around with a GoPro on our heads, that's maybe the most naturalistic movie stimulus, but that's not the one people, people use in the scanner.

And, and again, I don't think that's a problem. And I think we can capitalize on these. Stimuli that do are that are really powerful that are really engaging and make us feel [00:57:00] things that make us think things, uh, in ways that maybe just go pro footage from a street wouldn't so it's not argument to not use.

I mean, quite the contrary. I think it's just if we think of them as naturalistic, we're sort of doing, doing them a disservice in the sense that they are very So, um, yeah, Deliberate artistic products and, and we should think carefully about, you know, what they're, they were originally intended to do to their audience in order to assess, you know, if and whether, uh, and which of them is the most appropriate stimulus for our particular use case.

Benjamin James Kuper-Smith: Yeah, um, this might not be a hundred percent directly relevant to what you, I mean, it's, it's relevant to the whole conversation. Um, I recently had an interview with, um, Nahum Ulanovsky, who does really cool stuff with, uh, well, he does spatial navigation in bats. Um, so recording, you know, individual neurons as bats, bats are flying around and doing all sorts of stuff.

And, uh, we talked a little bit about, for example, he, you know, says that the whole natural versus controlled, it's not like one dimension, it's two dimensions and you can find these. Behaviors that are [00:58:00] highly controlled in the sense that animals do them naturally and repeatedly. Um, and that kind of stuff.

And that, yeah, it's not necessarily a compromise you have to make. It's more of a, uh, two dimensional space on which all kind of behavior fits. Um, yeah, again, it wasn't exactly a direct reference to what you just

Emily Finn: Yeah, no, that's I, I, I, I think it's interesting to think of it as, uh, not just as unit dimensional continuum, but, um, more. Multidimensional space.

Benjamin James Kuper-Smith: How do you, how do you get those, those, the films, for example? How do you choose them? Do you, you know, there's, there's, Millions of films out there, not, not counting everything you can find on YouTube. Uh, or TikTok or whatever. How do you select the right stimuli to be, I guess, in that sweet spot between Well, I don't know whether it needs to be in the sweet spot between, kind of, a natural video and controlled.

But, kind of, how do you Yeah, there's so many to choose from. How do you select the right one?[00:59:00] 

Emily Finn: Yeah, it's a good question and a hard question and the irony is that I'm actually not much of a movie person or like I, I have terrible taste in movies and I like haven't seen very many of them and so my students are always like making fun of me. Because I never have good suggestions but yeah I think It's a tough question.

I mean, I've been sort of on this soapbox for a few years now of like, you know, maybe let's replace at least some of the resting state acquisitions with movie acquisitions, because I just, I think that they're, you get more bang for your buck. They're more powerfully and flexibly analyzed. I think having that ground truth time course of what's actually on the screen or in people's ears at any given moment gives you a really nice handle into what's actually going on with the brain activity.

But of course. then the question becomes, well, which movie should we use? And it's, it's a tough one. And I think we're still in sort of early days of this work. And I, I would never pretend that there's sort of like one optimal stimulus or even like an optimal set of stimuli. I think, I mean, the, the cop out answer is it very much [01:00:00] depends on the research question, but, uh, you know, I think.

For example, like, so my lab is very interested in individual differences and how people interpret the same stimulus. Uh, and so one thing we've done is we've searched for sort of more independent films that people are less likely to have seen before. And we've searched for films that do kind of leave open a little bit of ambiguity as to what might be happening.

It's actually harder than you might think. So I spent a non trivial part of my postdoc, don't tell Peter, oh I think he would have been supportive, like literally watching indie films on, on YouTube and other uh, websites uh, to try to sort of like find stimuli that were doing what I wanted them to do in the sense of creating a story world that was It's interesting enough to be engaging, but still ambiguous enough to be plausibly interpreted in different ways by different people.

It turns out to be harder to find those than you might think. I think most directors and filmmakers are quite good at what they do, and they always have kind of some interpretation that they want you to get [01:01:00] out of it. And so even if it feels ambiguous, uh, there's often sort of less variability in what people ultimately come to than, than you might want there to be, at least with these professional and semi professionally produced films.

So that kind of leaves you with. a couple options. You can make your own films, which is difficult, although perhaps with generative AI getting less difficult. Yeah, I mean, I think you want to ideally, I think, select stimuli that are going to be, you want Engaging for people. Number one. I mean, I think that's part of the benefit of using naturalistic stimuli is it's actually more enjoyable for subjects than simply lying there and resting or usually even performing a more traditional task, and I think It is important to think carefully about the time varying features of the movie.

So I talked about, you know, luminance before being this like confound, so to speak. But you can actually pull out these feature time courses of not only low level things like luminance and loudness, but also Like mid level object categories, like is there a face on the screen? Is there a house on the screen?

Uh, you have [01:02:00] language, you know If you have a movie with dialogue in it and then you have these kind of high level features Like what is the emotional tone or is there a suspenseful moment and thing and usually for those you have to get kind of human writers to annotate them and people may or may not agree on those things but Like I think one practical piece of advice that i've given folks is when you're planning a movie study.

Like, if you know there's something that you're interested in, like maybe it is the the language or the dialogue, or maybe it's kind of that emotional tone, you can kind of try to pull out a lot of these features and just see to what extent they co vary across the film. You know, if you're really interested in studying Uh, these high level emotional changes, and you don't want that to be confounded by lighting changes, for example.

Like, you might want to look at that beforehand and kind of see, okay, if those two things are very strongly collinear, where, like, every time the movie enters this, like, dark phase, it literally gets darker, then You know, you can make argument that maybe that's not the best stimulus and you want to kind of be able to separate kind of these lower level things from these higher level features.

So that's just sort of very like practical piece of advice that I've given folks. But I [01:03:00] really think that at the end of the day, we have enough evidence that Having any movie on board for many, not all, but many use cases is better than just resting state that you can't really go to wrong. And there's not a super principled way, I would say at this particular moment to choose a film.

And so, you know, go with what you think is going to resonate with your participants, whether they're. Healthy college students or children or older adults. Um, ideally something people haven't seen before, because that can introduce another element of variability. If some of your subjects have seen it, or if you're taking a clip from a longer film.

Where some people have seen the whole film that could change, kind of, if some people are able to contextualize that clip in the context of a larger story. So, so there's sort of these experimental variables that you might want to think about. I have a paper with a former postdoc, Claire Grahl, where we kind of outline some of these practical considerations.

But, but I would say also just don't be afraid to pick something that you like and you think is compelling and you think is going to resonate with your audience because that's really at the end of the day, like, what we're using these for is to get people engaged and to kind of drive [01:04:00] their brain in these interesting ways.

Benjamin James Kuper-Smith: I suppose it shouldn't be too emotional that people get too many movement artifacts. I could imagine especially comedy could be, lead to a lot of like, head movement because of people laughing. I don't know whether you've had that, but

Emily Finn: I guess we haven't used stimuli that are like so funny that people are actually laughing in the scanner. Um, it's interesting. So there's been some work showing that, uh, especially in harder to scan populations like kids or patients, movies actually overall decrease head motion, uh, relative to rest at least.

Uh, the other big question that we've grappled with. I don't think anyone's published a comprehensive treatment of this, but is is whether head motion is correlated across people because that might not be ideal if you want to use some of these. inter subject correlation approaches. So like, if you and I are watching the same movie, do we actually, um, move our heads at the same time, right?

Because maybe there's some scary thing that jumped out, or maybe we're both laughing, and um, we've looked into this. We haven't published anything on it. It usually tends to [01:05:00] be less correlated than you might fear across people, but yeah, that's definitely something, or like, physiological signals, if everyone's kind of like holding their breath.

I mean, I think there's lots of things that you might want to try to, Measure, um, and record. But in general, it seems like for most populations, the data quality during movie watching is just as good, if not better than at rest. And that's in terms of things like head motion or arousal. People are more likely to stay awake.

I think during movies and physiological artifacts and stuff like that. So that's been our experience so far.

Benjamin James Kuper-Smith: Another kind of pretty big and obvious question, at least to me, is I think most people think that the more natural and normal, uh, the more everyday kind of Stuff people have to do and experience in a scanner the better but like you have to be able to analyze it And what am I going to do with people watching?

whatever film for 20 minutes, an hour, I don't know, probably depends. I suppose you could just repeat the stuff that's been developed [01:06:00] for Resting State. Um, I assume you can pretty much just apply that more or less one to one, but I'm curious, like, what's the, yeah, kind of, what do you do with all that? Okay, so now I have data of people watching film, what do I do with that?

I mean, you already mentioned you might have to get people to rate certain events, or, yeah, all that kind of stuff. So I'm just curious, like, how do you, how do you, how do you, what do you do with the data?

Emily Finn: So I think this is one of the coolest things about naturalistic data is they're very flexibly analyzed. So the same acquisition, you know, the same scan, you can really analyze it in lots of different ways. And so you can do all of the same functional connectivity analyses that you might be able to do from resting state, where you actually just sort of forget about the stimulus altogether.

But, you know, you're, you're kind of analyzing the, these fluctuations or the sort of intrinsic organization of the brain. Uh, you can. compare that across different movies if you want to, or you sort of like use the stimulus in some sense in that way, but you, you can just ignore it and do your sort of classic bread and butter functional connectivity analyses on movie data, but perhaps more interestingly, or like, you know, [01:07:00] as more of a point in favor of actually going to the trouble to use movies.

So you can I was sort of alluding to this before, but you can actually extract features from a movie and model them in a very similar way to you know how you would do it for a more simple task. So, um, movies do carry visual and auditory information that can be modeled. I mean, you're not controlling it, but you can model it so you can get kind of a luminance time course for how bright the movie is at different moments.

And you can take that as a regressor, basically back to your brain data. Um, Hope you see some some some changes and visual cortex. For example, if you're looking at luminance, you can do the same thing for auditory information. You can model these, uh, these more mid level features and, and recover, for example, like fusiform, uh, as faces are on screen or, uh, in response to the changes in faces when faces are appearing or disappearing.

You can treat movies like a task in the sense that, you know, you can model different features of, of the task and, and see, uh, which brain areas care about those features. [01:08:00] Again, you have less control over the features, so some of them might be collinear, and so you might Um, run into problems that way, but the benefit is that, you know, you're getting this really rich multimodal, I guess, if you have audio and visual information, you're getting multimodal, but then you're also just sort of getting the brain in this much more, um, Uh, yeah, naturalistic, I guess, for lack of a better word, but like, you know, you're not using these very stripped down stimuli.

You're sort of giving the brain, uh, input that's closer to what it might be experiencing in everyday life in the sense that there's a lot of stuff on screen all at once. You're kind of paying attention to different things. Uh, and so you can kind of use them, uh, use movies to, to confirm findings that people might have discovered in, um, uh, More controlled stimuli to say like, okay, once I now build up into this, um, screen that's messy and there's a face over here and there's a house over here and you know, there's movement and the luminance is changing all the time.

Like, do I still see, you know, some of the same regions that I might expect that cared about these isolated features and these low level tasks are those same areas caring about those same features once I sort of [01:09:00] build out to this more complicated situation. Um, and then you can also potentially discover new relationships between features and, and, and brain areas, or like the interaction between different features that might not have been possible with, uh, the lower level tasks.

So, so that type of approach really, I think, requires you to have a forward model of the movie, so to speak. So you have to know kind of what features you want to model, and you also have to know how to model those features, right? And that's. Maybe pretty straightforward in the sense of like, is there a face on screen or not, but it becomes more difficult in the sense of these higher level features of like emotional valence time courses or like suspense or anxiety or things like that.

And so it's very possible. And so you can get human annotations of listening. So that's one thing you can do. You can also take these. I think of as more data driven intersubject approaches. Uh, so this idea of intersubject correlation that was first introduced by Uri Hassan back in 2004 and, um, I think is an extremely powerful and elegant way of looking at the data.[01:10:00] 

So, in that approach, you're saying, I'm not necessarily, I don't really know what it is about the movie that the brain cares about, and I'm not going to assume that I can model it, but I'm, I am going to assume that there's some correspondence such that if you and I watch the same movie, and it's, you know, time locked in the same way.

Okay. That there's probably some shared processing going on, such that if I take the same region in my brain and correlate that activity time course with, you know, that same region in your brain. If I see that you and I are correlated in our activity as we're watching this movie, uh, I can infer that that region must be processing the movie in some way, right?

Because otherwise, why would it show? a correlation in its activity across two different brains at two different times and two different scans, etc. And so you can kind of use one person's brain as a model for another person's brain rather than modeling the movie directly. You're saying, okay, the movie is inducing some signature of activity that's going to be consistent across brains.

And so if I just look for what's consistent across brains, I can start to get an idea of which regions are responding to [01:11:00] the movie. Um, and so you can do that both in terms of space and in terms of time. So you can look for particular moments where people become synchronized. Um, you can look for particular brain regions where people tend to be synchronized over the whole movie.

Uh, and you can use that to, to try to understand something about group level processing. And you can also take that to an individual difference framework and ask, for example, like, are these two people more synchronized as they're watching the movie than these other two people, you know, and if so, what are those people share?

that these people don't. And then there's, you know, lots of newer approaches to where you can look at brain state changes and like event segmentation. So as you're kind of in one part of the film and then you kind of switch to another part, where do you see kind of state changes in the activity patterns in the brain?

Um, can you automatically infer those? Uh, you can take an encoding model approach where you're trying to learn mappings between the the features of the stimulus and, uh, what that does to the brain and actually predict, you know, in a brand new stimulus given a subject and [01:12:00] the stimulus, what would the different region time courses look like as a function of the known features of that stimulus?

And that kind of gives you a handle on, you know, if you can sort of discover the, the, the model that transfers stimulus features to brain activity, then you've learned something about, you know, what the brain is actually responding to in, in the stimulus, um, in this. Yeah,

Benjamin James Kuper-Smith: I have a very simple question about repetitions. So I guess one reason why the original task based experiments are so unrealistic is because you want, you need lots of repetitions to actually get to the effect. How do you deal with that if you have a, I mean, Okay, some visual stuff. I can, you know, I can clearly see how you can have lots of repetitions of certain visual features.

Be it seeing a face or something like that. But when it comes to, let's say, emotions or something like that, let's say you, you know, you're interested in betrayal. So [01:13:00] you show someone a film where someone is betrayed. There's probably gonna be one moment where that really happens in the film, right? And not a hundred.

I'm assuming you're not showing them the same 10 second clip over and over again. Uh, I'm curious, I mean, I would imagine that maybe with naturalistic stimuli, the effect is larger. People actually feel the betrayal more than maybe they might in some sort of random task. But, yeah, I'm curious, kind of, how do you, how do you get good statistical power when you, in my assumption at least, don't have hundreds or at least tens of repetitions?

Emily Finn: it's a good question, and I think it hits on this more. There's sort of one, a more general, I suppose. I won't say disadvantage. I'll say maybe caveat or thing we need to still work out is, um, you know, what actually happens in the brain as we watch the same movie a second time, right? There's sort of this paradox actually, and like film studies, uh, and, and, and literature and other, like media studies is like, we do actually [01:14:00] often, um, you know, Just in our everyday lives, we seek out the same media over and over, right?

Like there's maybe like a favorite movie that you have and you may have seen it, you know, many, many times or like, sometimes we return to the same book at different points in our lives and stuff like that. And so like, even though we know what's going to happen, there's still some, like. There's still some attraction or enjoyment that we get out of experiencing that piece of media a second or a third or, you know, so on, um, number of times, uh, I think it's, it's still sort of an open question, like, what happens, uh, what the effect of repetition of a movie clip is, for example, in the brain, like, you would, I think you've never really watched the same movie twice, right?

The second time around you have, Memory and anticipation effects, or maybe you're actually paying attention to slightly different things, uh, and, and this is something that people have started to look at this, um, in terms of what is sort of reliable across different presentations and what actually changes as a function of how many times you've seen that stimulus.

I think we have a little ways to go to sort of understand that better. I think it just sort of impacts the [01:15:00] you can draw. So, so in your example of like, if I'm interested in the neural response to betrayal, and I have only sort of one film clip and I show everybody that clip and I kind of see what happens at that particular moment where you learn that the character has been betrayed, I think.

There have been, I mean, we've done some of these and there's been other studies showing that you, you know, you can get interesting information from just sort of that single trial, so to speak, when you're sort of looking across subjects. But I think in that case, you know, you wouldn't necessarily be able to claim that that would generalize to any instance of betrayal.

So, like, I think one kind of happy medium that we've been thinking about, I mean, this, I, I acknowledge that this is challenging, but I think it's doable. It's like, you know, if you can build out sort of a. a suite of clips or a set of clips or something that all kind of have some moment that feels similar in some like abstract sense.

It's almost like you want to abstract away from the details of the individual clip. Um, but if you can get that betrayal moment across like, let's say five stimuli, um, that are all different in terms [01:16:00] of exactly what's happening and who the characters are, but then, you know, you, you can have sort of.

Slightly more of a trial like structure, um, but you're kind of, you're moving away from that exact particular clip and you're trying to generalize across different experiences. I think something like this would also be useful for longitudinal designs, you know, if you want to scan the same people across development or across, um, some other, probably like a treatment for a mental illness or something like that.

Like maybe the ideal thing is not to have the exact same clip. Maybe you want to have one thing. Of, you know, one identical clip for sort of test retest reliability, etc. But then it would be nice if we could move to a model where we have sort of if we have a good parameter ization of these stimuli and what's actually happening and what we care about, like, in theory, we could either find or generate.

sets of stimuli that are all sort of different, like in an item wise sense, but that all kind of tap into a similar process, and then, um, we're kind of generalizing at that second order, [01:17:00] more abstract level, rather than, you know, the exact, like, audiovisual details of that particular clip at that particular moment.

I don't know if that makes sense, but

Benjamin James Kuper-Smith: Yeah, it does. But in a way, I guess one, I guess one problem maybe is that, I mean, again, I, maybe this isn't a problem. It just seems to me like that is that, you know, in these artificial tasks, I guess the reason, I mean, not the reason artificial, but you know, it doesn't take that long maybe until you can experience a broad array of different emotions you might feel included or excluded or something like that.

Whereas in a film, it just. It just takes time, right? It unfolds, at least it feels to me, often more over time. Uh, it seems to me like if you have some very simple, like, game you made up, you can maybe get, like, every few seconds, like, a little bit of it. Whereas in a film, it might take five minutes before you even really understand what the relationship is between the people that you're seeing.

Um, yeah, I'm just curious whether [01:18:00] that just inherently limits the how many repetitions you can do of something? Or is it just, why am I just thinking maybe also just from the entirely wrong perspective about this?

Emily Finn: No, I, I, I mean, I think you're right. Like, I, I wouldn't necessarily argue that movies need to subsume All tasks, right? I think they can be, they can be powerful for kind of discovery science or like more data driven, like, like, okay, I mean, I guess there's sort of multiple answers to this point that I could give.

But yeah, so I, and like, I think, I mean, you can also use like quite short clips, right? So that's like a way, but then you're kind of starting to approach this more, um, traditional task based trial based structure. And then at that point, it's kind of like, what's the point? Like, I, I do think. Okay. I do think there's something to using slightly longer clips, like, I don't know.

Yeah. In the 5 to like, 10 minute range, let's say, or even beyond that, that does let your participant kind of immerse themselves in this narrative that could then make that moment ultimately more powerful [01:19:00] and ultimately more similar to what we might experience in our own lives. Right? So, like, I do think and I, I hope I'm clear.

I mean, I do think it's. It's important and good to kind of use movies in conjunction with maybe a more stripped down, controlled, trial based paradigm. And hopefully there's some similarities between, for example, being like betrayed in a gambling game versus like watching a betrayal on screen. And, and, but that's an empirical question of like, you know, how much, how much similarity is there between those two things?

But if we Uh, if we use movies as a tool to get participants really kind of like invested in the story, I think that does lead to more powerful signals and perhaps signals that are more akin to what we might experience in our own lives. I think the benefit of movies also is that, yes, it's nice and it's good to sort of Go in with a hypothesis.

And obviously, I alluded to this when talking about picking a movie is like, ideally, you have some sense of, you know, the effect you want the movie to have on your [01:20:00] participants. And then you can model particular things about the movie that that might induce or relate to that effect. But the nice thing about movies is they also give you a chance to kind of let the data tell you what's important in the sense that if you see like a strong spike and correlated activity at a particular moment that you didn't anticipate.

Kate. You know, doing that, you can kind of go back to the movie and say, well, what was happening at that point? Like, is there something here that I wasn't thinking to model, but maybe is important or is a feature that the brain is leveraging or reacting to? Um, of course then you probably have more work to do to then kind of build another context in which you are directly manipulating that thing so that it's not just a big, like reverse inference or storytelling game, but I do think movies give you that potential of.

Letting the brain tell you what it cares about, as opposed to you as the experimenter kind of building a task in a way that you know exactly what you want to model. And again, that's very powerful. Like, I think we need both of these approaches, but I think, I think that the [01:21:00] benefit of movies over these more controlled tasks is it gives you a certain set of extra degrees of freedom, so to speak, such that you can go in with something that you might want to model and something that you might want to test, but it also leaves room For new moments or new features to present themselves that you can get from looking at the brain activity that you wouldn't have necessarily hypothesized in advance.

And therefore, you might not have baked into your controlled task. Um, uh, yeah,

Benjamin James Kuper-Smith: yeah, definitely. Yeah, yeah. Yeah, I mean, to some extent, I guess all of my questions kind of revolve, or are supposed to revolve around the question of kind of, for what kind of question or type of signs are more, is more useful, and when is something else more useful? Um,

Emily Finn: I think like, yeah, just to. Make one more point of this, again, like I, I don't think we should subsume like all of fMRI research. I do, I do think like one point that I've been trying to advance, like what one place where I think they should play more of a role than they do right now is in [01:22:00] these large scale consortium studies.

So things like the Human Connectome Project or even like UK Biobank or The ABCD study, like these are really powerful kind of population imaging efforts where people are going to huge efforts and, and spending a lot of money to like acquire a lot of data on a lot of individuals. And that can be very powerful.

Uh, but I think in my view, it's a little bit unfortunate that a lot of the paradigms that get included in these studies are kind of these more traditional tasks that, that, that maybe, um, I mean, we've gotten a lot of mileage out of those, but there's also a lot of evidence now that those are not actually the best for revealing stable and meaningful inter individual differences.

Um, and those types of studies sort of by definition are kind of trying to do discovery science, right? It's like saying, let's collect a lot of data on a lot of people, um, and then we can kind of mine that in different ways. It's not like you're going into one of those studies with a very specific hypothesis that you want to test.

The idea is just to get a lot of data and kind of mine in different ways. [01:23:00] And so I think movies for that application could be a more powerful alternative to resting state because you can do all the same things that you can do with resting state with movies. But you can also do all of these additional things where you can leverage the fact that you have that known time course to understand, you know.

Why different brains are sort of reacting, or maybe not why, but, but like sort of what the reactivity of different brains looks like to different features in the movie, and you can just kind of, yeah, analyze that data much more flexibly, given that you have this, this ground truth time course of what was happening on screen.

So I think that's one place where maybe they're underutilized relative to Resting State, for example, and should maybe get more. airtime in the scanner. Um, but again, I think they're most powerful probably in conjunction with, with more targeted and kind of hypothesis driven tasks to see if we can get a handle on, on the processes that we think we're studying in both of these contexts, you know, to, to see how, how much they, they relate across contexts.[01:24:00] 

Benjamin James Kuper-Smith: Yeah, so as I mean, as I, as I told you, uh, before we started recording, I, um, you know, I interviewed Peter Vannettini a few, few weeks, a month ago, and after we stopped recording, I mentioned that I was going to interview you, even though at the time I hadn't even, I somehow missed that both of you, that you'd worked for him or with him.

And then when he said, you know, that you worked in his lab for a few years, I was just curious whether there's anything he, uh, he said he thought I should ask you about. And he gave some very good questions that all went way above my head and were way too technical for my analysis. So I'm not going to ask any of those.

But he did mention something I thought was very interesting, uh, just as a kind of almost an aside, because he basically said you were the best writer. Basically, he's known as a, I guess, as a collaborator. He basically said that. Uh, I mean, something to the effect of, you know, usually you'd send him a manuscript for comments and he'd just say like, yeah, looks good, he wants to submit it.

And so I was [01:25:00] very curious about that because writing is something that, let's say, I want to be good at. Yeah, it's a very broad question, but kind of how, how, how do you do that?

Emily Finn: Well, first of all, that's a really nice compliment and vote of confidence from Peter, so I appreciate that. Uh, I've always loved to read and write. I mean, I've always been kind of inclined towards, uh, yeah, literature and journalism and stuff like that. So I do think that helped, uh, and I actually think scientific writing should be less dissimilar to good journalism or even good fiction than it is.

Yeah, I mean, I think there's a certain amount of conventions that we have to follow and that are useful, but I also think it's, it's useful to kind of draw inspiration from. Other types of expository writing, like long form journalism, kind of see how people build up arguments and in long form articles like that.

It's not going to be directly applicable to the structure of a scientific paper. But I think at the end of the day, I mean, a good paper, good grant kind of tells a story, right? I mean, it's, it's, It's a bit of a cliché, but it [01:26:00] is for a reason, and you know, you have to keep your reader in mind, and someone's out there reading this, and try to make it pleasant for them to read.

I mean, I think that your ideas will have a much bigger impact that way. And as to how to do that, I mean, I, I spend a lot of time on my writing. It's not something that comes quickly, um, but I think I like the metaphor. Again, this is probably overused, but you know, writing is more like, like sculpting than painting, in the sense that Like, what can you subtract?

Like, what can you take away to make it better? So you often start with just kind of like these messy drafts readers getting all the ideas on the paper. And then it's kind of like, okay, how can you streamline this? What can you cut? And like at every level. So it's like within a given sentence, can you like cut a few words or is there a single word that can do.

What you're doing with a three or four word phrase, right? And even those small changes, those kind of add up and they make it easier and just more pleasant for people to read. And then also on a larger scale, like, why is this [01:27:00] paragraph here? Like, what is it doing? Is it just here because you want to show off that you know this literature, but it's not actually super central to the argument that you're trying to make.

And just kind of taking a very critical eye to every word, every sentence, every paragraph that's on the page, trying to make sure it's doing work and not just there as kind of a. Okay. Oh, here's another sentence or here's another thing I could bring in. And it's hard. I like, I, yeah, I mean, I think I still, I sometimes go back to like my writing from a couple of years ago.

And even that I, you know, I'm sort of dissatisfied with it, but, um, I really think it's writing with your reader in mind and, and just trying to, yeah, like be as concise as possible and what you're trying to say and varying your sentence structure is sort of a classic piece of advice, but it can really help, uh, having some shorter sentences in there.

And, but it's, yeah. It's a labor of love and I'm probably too much of a perfectionist with it. And it's still not, it's never going to be perfect. And, um, yeah, but I think it's a really important part of what we do is, is communicating it in, in written form. So I appreciate that Peter said that. I

Benjamin James Kuper-Smith: Yeah, I really like [01:28:00] the point where you said that if one word can do the job of four words, something like that, because that's something that I You know, I do all the time basically, even rewriting, it's just like, where am I spending, you know, where is someone spending mentally five seconds on this when they could be spending one second on it?

And I feel like that's, I mean, it's something I do all the time, but it's never something I would have, I don't think I would have come up with it. But when you mentioned it, I was like, oh yeah, that is, that is a big part of clarity through just eliminating the stuff that doesn't need to be there or can be said shorter.

Yeah. Do you, but I guess, I mean, did you do lots of writing and other things before? I mean, I guess you mentioned a bit of journalism or the, at MIT, the kind of press releases, that kind of stuff. Do you think that helps by just having kind of a contrast to what it isn't, to just knowing more specifically what you're trying to do with this research paper?

Or does it just help because you've just thought more about language and how that [01:29:00] works?

Emily Finn: think it's helpful to have had practice and experience with other types of writing beyond science writing. I mean, I think most people that start a PhD program have probably written things in their lives, like, I don't know, I guess, like, research papers for school or other, um, essays and stuff. So people will sort of come in with varying degrees of experience with that.

I mean, I, I did always, like, you know, I was, like, growing up, I would, like, write little stories as a kid and illustrate them. And so I guess I've always been sort of attracted to the literary arts, so to speak. But I definitely think it's something that can be developed and cultivated, like, even once you're in the field and just kind of forcing, like, something I tell my students is when you're reading papers, which hopefully you are sort of actually reading from start to finish, not all papers, some papers, you will skim some people's you're reading in more depth when you find a paper that you think is really well written, just kind of file it away.

And it doesn't even have to necessarily be for the content of the paper. But if you find a paper, that's actually. A joy to read as opposed to a bit of a chore to read, you know, [01:30:00] file that away somewhere and, and, and come back to it and see kind of, you know, how did the authors craft that just on a very kind of, um, like, yeah, craft over content.

So, like, if you just break down the introduction into, like, what each paragraph is, is doing and, and also the, the style of language that they're using, uh, can you, yeah, just start to deduce, you know, what makes some papers. It's more pleasurable to read than other papers, and use that when you're crafting your own papers to kind of, yeah, do your readers a favor and make it interesting for them.

And people all different styles, like I don't think there is one formula, but I think if you can kind of train yourself to start recognizing it when you see it and you kind of compile some examples for yourself, you can kind of start to extract some, some principles of what you think makes a good paper and then try to echo that in your own writing.

Hmm,

Benjamin James Kuper-Smith: just briefly on the, on the topic of papers that I enjoy to read, I guess I'll [01:31:00] recapitulate that question now, because uh, the first of my recurring questions is what's a book or paper that more people should read? Can be famous, non famous, old or new? I mean, I don't know whether you had a, whether this was or whether you're going to be your answer, but I'd also be curious if you have a paper that is a joy to read.

Emily Finn: that's a good question. I probably should have been more prepared for. I mean, the paper that actually immediately came to mind when you told me you were going to pose this question over email is this paper from 2018. I just have it right here. So I don't butcher the title, but it's called. The Reliability Paradox, Why Robust Cognitive Tasks Do Not Produce Reliable Individual Differences.

The first author is Craig Hedge. It was published, um, yeah, in 2018 in Behavioral Research. I, I love this paper. Uh, so this paper is actually very elegantly making the point from both a theoretical and an empirical perspective. I sort of I [01:32:00] vaguely alluded to this, I think, earlier in the conversation, but it's basically making the point that the sort of bread and butter tasks of cognitive psychology became that way, like became bread and butter, because they produce very reliable effects in basically almost every subject that you test.

But that very feature makes them poorly suited for discovering meaningful individual differences. So, if, if you have a task that, um, It's just very reliable in the sense that it kind of produces very stable behavioral effects, both within and across people. That task is, is, is well suited to understanding something generally about how humans or human brains perform that task, but it's not well suited to drawing out.

These meaningful differences across people because by death by some

Benjamin James Kuper-Smith: Well, the whole point is,

Emily Finn: Yeah. I mean, some of the traditional definitions of reliability is like, well, you want to see similar effects in every subject that you run. [01:33:00] Right? And so that was true. And, like, the earlier days of experimental psychology was also true in fMRI, right?

Like, a lot of the early papers were like, well, like, you want to see motor cortex lighting up in every individual that you test, or at least most individuals that you test and that's how you can kind of trust. Um, like, I think the original original, um. Fusiform base area paper out of Nancy Kanwisher's lab, like they were actually showing some individual subject data, they were sort of presenting results and like, you know, they discovered the fusiform in, you know, nine out of the 12 people they tested.

I can't remember the exact ratio, but it was like, that's a strength, you know, um, when you're trying to prove something general about the population. However, um, it's a weakness when the thing that you actually care about is what is meaningfully different between people. And so you're kind of shooting yourself in the foot.

It's By using these like tried and true paradigms when you're trying to go after this, this, this variability across people. And so we kind of need to take a step back and think about designing new tasks and new paradigms that. don't necessarily produce the same effect in everybody, [01:34:00] but rather produce different effects in everybody in a way that's still sort of stable and reliable within each person, right?

And that's challenging, but I think that's the way that we need to move towards if we're going to get traction on some of these, these individual difference questions. And so this paper, The Reliability Paradox, just kind of lays out this, the whole principle behind this in a very, I think, clear and elegant way, um, just like why it kind of has to be true theoretically.

And then they also do some empirical analyses of, um, some sort of classically used tasks like flanker tasks, stroop tasks, stop signal tasks. And they kind of show that, you know, the very feature that makes these tasks desirable for sort of like general cognitive psychology research makes them undesirable for individual difference research.

So that's probably been my favorite paper in the last, yeah, several years and I cite it all the time. And I think it's, it's quite well written as well.

Benjamin James Kuper-Smith: Nice. Uh, second recurring question is something you wish you'd learned [01:35:00] sooner. This can be from your, from your work life, from your private life, doesn't really, I don't really care, just something that you think that, uh, your life, your life would have maybe been a little bit better if you, if you'd figured that out a little bit sooner, and maybe how you figured it out or what you did about it.

Emily Finn: Yeah, I think actually maybe this, this gets back to a second. I have a book recommendation. I actually just finished this book over the weekend. So this one is quite fresh, but there's a book called, um, I think it's called 4000 Weeks Time Management for Immortals and it's by Oliver Berkman. It's kind of like, Existential self help book, uh, basically making the point that, you know, we, uh, we can sort of bend over backwards and try to be these productivity machines and fit everything in at work and at home and just, you know, sort of kind of just try to.

Over optimize and maximize our time. But at the end of the day, we all have finite time here on this earth. And we're never, you [01:36:00] know, to the extent that we get more done, there's just going to be more to do. And it's kind of this never ending cycle. And, um, I, so I'm a parent now and I'm also a PI and I also like to sometimes like do things that are.

Not work. Uh, and, and it's challenging. And so I think the ideas in that book, I mean, it's not the first time that I had encountered some of them, but I think he lays it out in a particularly kind of, um, engaging, compelling, often funny, often kind of tongue in cheek way, uh, and, and I guess the, the point that I'm trying to make is like, I, I guess I, I wish that maybe I had been more earlier on, like a little less consumed with like, am I, sorry, let me back up.

This is like a hard thing to answer. I think, um, one thing that's challenging about a career in academia is you kind of, being your own boss sounds great and it often is, but it's challenging in the sense of you have to kind of really manage yourself and you have to do a lot of work. I often call it meta work.

It's like [01:37:00] work about work. So it's like you're thinking about Not only work, but like, if you're doing enough work, am I doing too much? Am I doing too little? Am I focusing on the right things? You know, it's like you have this infinite. Not infinite, but you have a lot of flexibility and a lot of freedom as an academic to sort of work on things that you want to work on and choose, you know, which activities are going to do and which things are going to say yes to.

And that's great. But it also, yeah, it creates this level of sort of extra management that you have to do to try to understand, you know, if you're actually using your time and in the ways that you want to be, and you don't have a boss telling you, like, prioritize this, not that, etc. Um, and so, um, I think, uh, that book really, you know, crystallized a lot of the feelings that maybe I was having and maybe other folks have too, of like, it's just sort of getting obsessed with how you're using your time in any given day and whether you're using it optimally and, and, and, and not just at work, but, but, you know, in your personal life as well.

And so I, I, I guess. I wish that maybe earlier I had [01:38:00] recognized this a bit better and recognize that there's always going to be more to do. And, you know, you're kind of always going to feel like there's infinite other things that you could be doing. But the best that you can do is just, you know, pick a few things that are important to you at any one time.

In terms of work, but also in terms of personal life and just kind of focus on those things and recognize that life has chapters, right. And just because I'm not running marathons right now, it doesn't mean I can't do it later or, you know, like something like that. And like to everything it's its own season.

And, um, just to kind of stress out a little bit less about like the productivity geek culture and just like maximizing every moment

Benjamin James Kuper-Smith: Yeah, I mean that, that book, you're not the first to mention it. I think Rachel Bader mentioned it,

Emily Finn: Oh, cool.

Benjamin James Kuper-Smith: and yeah, I also read it I think a year ago or something, and I want to reread it again because it's, uh, reread it again, yeah, I want to reread it. Speaking about when you can cut words, yeah, because it does have, yeah, I agree, as you said, this [01:39:00] kind of mixture of putting things in perspective, being weirdly calming at the same time as kind of telling you things you, you know, but don't really want to hear.

Emily Finn: Totally.

Benjamin James Kuper-Smith: Um. So yeah, well, I'll, you know, with all the other stuff I mentioned, I'll, you know, put a link to that in the description. Um, I would definitely recommend it. Um, final question. Any advice for PhD students or postdocs? So people like me who are kind of at that transitory period between, you know, finishing a PhD to having just started a postdoc.

Emily Finn: Yeah, I think specifically with respect to that period when you're kind of moving between PhD and postdoc, I think that's a really fun time. Maybe I'm not the only PI to have said this. I mean, I think when you're coming out of your PhD, you have a lot of skills and you have a good handle on the landscape of the field.

And so when you're starting your postdoc, you know, you're really kind of able to just go full steam ahead. And yeah, you already kind of have the skills on board and you can start [01:40:00] to, you know, make more rapid progress. At the same time, postdoc can be stressful because it's. It's typically supposed to be shorter than PhD and then you're worrying about the next step basically almost as soon as you start your postdoc so that part is hard but I would say like one thing that I think it's really nice and important to do as a postdoc is and I didn't ask you exactly where you're planning to postdoc and how that relates to what you did in your PhD but I do think postdoc like you have a lot of independence but it's also nice and that it's your career.

It's kind of your last chance to get that direct training or mentoring from somebody. And I think, yeah, like, like you want to establish independence, but you also want to take advantage of that, you know, being basically the last time in your life that you're going to have this like formal mentor who is in charge of you.

And like, by definition needs to care about you and what you're doing. Um, and, and just to kind of like lean into that and make sure like your, your. Taking that advice on board and just kind of like maximizing that opportunity. I think postdoc is also a nice time to force yourself to get some experience.

And maybe [01:41:00] this is already what you're going to do. I think this is what many people try to do, but it's like, you kind of want to. Add something to your toolkit. That's a bit different than what you did as a PhD student, because that's going to make you a stronger scientists moving forward. Like, I think, actually, when I joined Peter's lab, Peter's lab is fantastic.

And in many ways, Peter Panettini was similar to my PhD mentor, Todd Constable, in that he's an MR physicist and he runs sort of a methods focused lab, but, but with still with very strong neuroscience applications. Uh, and so I actually think it would have been easy for me, honestly, in Peter's lab to just kind of To a lot of the same stuff that I've been doing as a PhD student and I think that's very tempting for folks at that stage because it's like, as I just said, you have these skills and you kind of hit the ground running and you can start to crank out projects and you kind of like be really productive.

And there's a, there's a seductiveness to that, which, which, which makes a lot of sense, um, because our field really values productivity and. Fast, but at the same time, I think, like, one thing I did as a postdoc, almost by accident, but I think it was actually great. It was like, I kind of [01:42:00] ended up doing this whole side project about layer specific fMRI, which is not something that I had done as a PhD student.

It kind of gave me that exposure to a new set of methods and a new way of thinking about things. Um, I think in many ways, like if you're planning to join a lab that's more different from what you did as a PhD student, this will come more naturally. But I think it can be hard to kind of slow yourself down and kind of force yourself to to learn a new set of skills because you do have all this momentum propelling you and like, whatever direction you're coming out from your PhD with.

But I, I think that, yeah, kind of like forcing yourself to to do something a bit different where there's going to be more of a learning curve for you. And it might take a bit longer to like, for that project to come to fruition, but it can be. A really great thing to do to set yourself up to to be in your own independent career, because now you kind of have this extra dimension that you can bring to problems.

That's, you know, going to make you sort of more unique from what either your. Grad or your postdoc lab are doing and it kind of, you're, you kind of become this, you know, you're like more of the, some of your, more than some of your parts [01:43:00] almost because you can kind of see connections and, and, and do stuff in a way that if you would just kind of stayed in the narrow lane that you'd acquired your PhD, like, sure, maybe you'd have published a couple more papers, but, you know, in the, in the grand scheme of things, I think postdoc is a nice phase to kind of force yourself to, to do something a little bit outside your comfort zone because you do kind of have that protected time and space to do that a little bit.

Okay.

Benjamin James Kuper-Smith: Okay, great. So learn something new. By taking advantage that someone, for the last time in your life, is contractually obliged to care

Emily Finn: Yep.

Benjamin James Kuper-Smith: That's the, okay. Great, well then, thank you very much.

Emily Finn: It's a lot of fun. I hope, hope it made some sense. Thanks for having me.

Benjamin James Kuper-Smith: I think so. Thank you.